I'm trying to do LASSO in R with the package glmpath. However, I'm not sure if I am using the accompanying prediction function predict.glmpath() correctly. Suppose I fit some regularized binomial regression model like so:
fit <- glmpath(x = data$x, y=data$y, family=binomial)
Then I can use predict.glmpath() to estimate the value of the response variable $y$ at $x$ for varying values of $\lambda$ through
pred <- predict.glmpath(fit, newx = x, mode="lambda", s=seq(0,10,1),type="response")
However, in the help file it can be seen that there is also an option newy. How should one interpret the result when calling predict.glmpath() with newy = some.y?
[Edit] An additional question came to mind:
The option type can have the following values, according to the help file:
description in help file
"response" the estimated responses are returned
"loglik" the log-likelihoods are returned
"coefficients" the coefficients are returned. The coefficients for the initial input variables are returned (rather than the standardized coefficients)
"link"(default) the linear predictors are returned
However, to which linear predictors and coefficients are they referring to? Surely not those of the original model?
